IDENTIFICATION AND ESTIMATION OF A LABOUR MARKET MODEL FOR THE
TRADEABLES SECTOR: THE GREEK CASE
Costas Milas
Department of Economics
University of Warwick
Coventry CV4 7AL,UK
and
Jesús G. Otero
Department of Economics
University of Warwick
Coventry CV4 7AL,UK
March 1999
Abstract. This paper derives a theoretical labour market model for the tradeables
sector of a small open economy. Using Greek manufacturing data and applying
multivariate cointegrating techniques, two cointegrating vectors are estimated based on
the a priori restrictions provided by the theoretical model; a labour demand and a real
exchange rate equation, respectively. The short-run estimates of the model suggest that
labour decisions not only depend upon past disequilibria in the labour market, but also
on the discrepancy between the real exchange rate and its implied long-run equilibrium
relationship, that is, the magnitude of the real exchange rate misalignment.
Keywords: Labour market, tradeable goods, cointegration, identifying restrictions,
Greece.
JEL classification: C32, C51, C52, E24, O52.
1
1. Introduction
Based on a two-level Constant Elasticity of Substitution (CES) production
function, this paper introduces a theoretical labour market model for the tradeables
sector (proxied by the manufacturing sector) of a small open economy. Using quarterly
data for Greece, the model is estimated by the Johansen (1988, 1995) multivariate
cointegrating framework. The advantage of introducing prior economic information in
Johansen’s cointegrating approach is discussed in Wickens (1996), who points out that
researchers often fail to give an economic interpretation to their estimated
cointegrating vectors when they ignore economic theory.
Existing labour market models (see e.g. Layard et al., 1991) have mainly
focused on the interrelationships among labour variables within the context of a CobbDouglas production function. Recently, using a theoretical two-level CES production
function (see Bruno and Sachs, 1985; Alogoskoufis, 1990), Milas (1997) estimated a
labour market model of the Greek economy. In the current paper, we extend the work
in Milas (1997), by explicitly introducing in the theoretical model a long-run real
exchange rate equation, as a measure of price competitiveness. The importance of this
new feature, is that it allows for the effect of open economy issues upon employment
and output levels in the tradeables sector. Further, we use quarterly data as opposed to
annual data in Milas (1997), which allows the introduction of richer dynamics into the
empirical analysis.
Our findings support the existence of two cointegrating vectors which are
identified using the a priori economic information provided by the theoretical model.
The first cointegrating vector can be interpreted as a long-run labour demand equation,
which depends upon output and real wages. The second vector can be thought of as a
2
long-run real exchange rate equation, depending on real wages and international oil
prices. The short-run dynamics of the model suggest that labour decisions in the
tradeable sector not only depend upon past disequilibria in the labour market, but also
on the discrepancy between the real exchange rate and its implied long-run equilibrium
relationship.
The organisation of the paper is as follows. Section 2 introduces the theoretical
labour market model for the tradeables sector. Section 3 is devoted to the key
empirical issues: the properties of the data, the cointegration method and results, the
implied long-run equilibrium relationships, and the short-run dynamics of the model.
Section 4 offers some concluding remarks.
2. The theoretical model
Starting from a two-level CES production function in the tradeables (T) sector,
which is separable into capital (kT), labour (lT), and imported oil (oT), and assuming
perfect competition, Bruno and Sachs (1985) and Alogoskoufis (1990) have derived
the following log-linear equations for gross output (qT), oil (oT), and employment (lT),
respectively:
qT = π1vT + (1 − π1 )oT ,
(1)
oT = qT − ρ( pO − pT ),
(2)
lT = vT − σ( w − pT ) +
σ
(qT − vT ),
ρ
(3)
where vT is value added in the tradeables sector, ρ is the elasticity of substitution
between vT and oT, σ is the elasticity of substitution between kT and lT, pO − pT is the
relative price of imported oil, w − pT refers to real product wages, and π1 is the share of
3
value added in gross output (0 <π1< 1). All variables are in logs. Substituting (2) and
(1) in (3) gives the long-run labour demand equation in the traded goods sector:
 1 − π1 
( pO − pT ).
lT = vT − σ(w − pT ) − σ
π
 1 
(4)
According to (4), labour demand in the tradeable sector (lT) is a function of output
(vT), tradeable wages (w – pT) and the relative price of imported oil (pO – pT).
As for the non-traded goods, we assume that these are primarily services, with
production being labour intensive. Thus, we write in log-linear form:
vN = l N ,
(5)
where vN and lN refer to non-tradeable output and employment, respectively.
Assuming profit maximising monopolistic competitive firms in the sector, we
get optimal prices as a mark-up on unit labour costs:
pVN = w + µ,
(6)
where pVN refers to the value added deflator in the non-tradeable sector and µ is the
constant mark-up.
The log of domestic prices is given by:
p = τpVT + (1 − τ) pVN ,
(7)
where τ is the share of tradeables and pVT is the value added deflator in the tradeable
sector.
From (6) and the identity
pT ≡ π1 pVT + (1 − π1 ) pO ,
(8)
we write domestic prices as follows:
 1 − π1 
( pO − pT ) + (1 − τ)µ.
p = τpT + (1 − τ) w − τ
 π1 
(9)
4
Following Edwards (1989), competitiveness is defined as the relative price of
tradeables to the price of domestic output, that is, pT − p. Rearranging (9) in terms of
pT − p, we get:
 1 − π1 
( pO − pT ) − (1 − τ)µ.
pT − p = −(1 − τ)(w − pT ) + τ
 π1 
(10)
According to (10), competitiveness (pT – p) is a function of tradeable wages
(w – pT) and the relative price of imported oil (pO – pT).1 If an estimate of π1 is available
(e.g. from previous studies), then equations (4) and (10) can be simplified to:
lT = vT − σ ( w − pT ) − σ ( pO − pT ) *
(11)
pT − p = − (1 − τ)( w − pT ) + τ( pO − pT ) * ,
(12)
 1 − π1 
( pO − pT ) * = 
 ( pO − pT ) .
 π1 
(13)
and
where
Equations (11) and (12) constitute a set of long-run structural equations for
labour demand and competitiveness. Equation (11) assumes a unit coefficient on
output (vT), and equal effects from tradeable wages (w – pT) and the relative price of
imported oil (pO – pT)* on employment (lT). Equation (12), on the other hand, assumes
that the coefficient on tradeable wages (w – pT) is equal to the coefficient on the
relative price of imported oil (pO – pT)* minus one. The validity of these restrictions can
be tested in the cointegrating space. Furthermore, the theoretical structure of the
model allows us to derive estimates of some structural parameters of the economy,
Here, we assume that the mark-up parameter (µ) is constant. A future approach would be to model µ
as a function of demand fluctuations, that is, µ = − γ(y − yF), where y and yF refer to output and full
employment output, respectively, and γ >(< 0) if the price elasticity of demand is increasing
(decreasing) in the level of activity (Bean, 1994).
1
5
namely the elasticity of substitution between capital and labour (i.e. σ), and the share
of tradeables in total output (i.e. τ).
3. The empirical model
Our model uses a set of p = 5 variables, z = [lT, (pT – p), (w – pT), vT, (pO –
pT)*]′, for quarterly seasonally unadjusted Greek manufacturing data over the period
1963q2-1997q2.2 Labour demand, lT, refers to manufacturing employment. Value
added, vT, refers to manufacturing output. Tradeable earnings, w, refer to average
earnings in manufacturing. Assuming that the Greek manufacturing sector is fully open
to international competition and hence is a price taker, we proxy pT by: pT = pT* + e,
where pT* is the dollar unit value index of industrial countries imports and e is the
average drachmae/dollar exchange rate. Competitiveness is proxied by the international
price of tradeables relative to domestic prices: pT − p = pT* + e − p, where p refers to
domestic consumer prices less food and rent.3 Competitiveness can be interpreted as
the real exchange rate. In this case, an increase (decrease) in pT − p is equivalent to a
real depreciation (appreciation) of the domestic currency. The price of oil, pO, is
proxied by international oil prices (for a full description of the series see the Data
Appendix). Firms are assumed to take pO – pT as given, so that we can impose
exogeneity of this variable for the rest of the system. In order to test the validity of the
restrictions implied by equations (11) and (12) above, we assume that π1 is equal to
2
The choice of seasonally unadjusted data is in line with Ericsson et al. (1994), who show that the use
of seasonally adjusted observations may affect not only the short-run dynamics of the system, but also
the week exogeneity status of the variables (see also Chapter 15 in Hendry, 1995).
3
In terms of relative price competitiveness in international markets, consumer prices less food and
rent seem to be a more appropriate index compared to the more general index of consumer prices.
6
60.2% (see Milas, 1997). Therefore, using (13), (pO – pT)* is constructed as
( pO − pT ) * = 0.66 ( pO − pT ) .
Using the notation described in Johansen (1988, 1995), we write a p
dimensional Vector Error Correction Model (VECM) as follows:
∆zt = Γ1∆zt −1 +...+ Γk −1 ∆zt − k +1 + Πzt −1 + µ + ΨDt + ε t , t = 1,..., T
(14)
where ∆ is the first difference operator, zt is the set of I(1) variables discussed above,
ε t ~ niid (0, Σ) , µ is a drift parameter, Dt is a vector of non-stochastic variables, and Π
is a (p x p) matrix of the form Π = αβ′ where α and β are both (p x r) matrices of full
rank, with β containing the r cointegrating relationships and α carrying the
corresponding loadings in each of the r vectors. Assuming exogeneity of (pO – pT)*
enables us to reduce the number of variables to be modelled from five to four.
However, we allow (pO – pT)* to enter the cointegrating space, which implies that input
prices have a long-run impact on the system. The Dt vector includes (i) three centred
seasonal dummies and (ii) a dummy variable, DEVAL, which takes the value 1 in
1983q1 and 1985q4; 0 otherwise. The variable DEVAL captures the step devaluations
of the domestic currency by 15.5% in January 1983 and 15% in October 1985,
respectively.4
Preliminary analysis of the statistical properties of the data using the
Augmented Dickey-Fuller (ADF) tests suggested that all series are I(1) in levels and I(0)
in first differences.5 Table 1 reports the diagnostic tests for the levels of the four equations
and the system in (14), using a lag length of k = 5.6 Cointegration results are shown in
4
Both devaluations were underpinned by tight stabilisation economic programmes as discussed for
example in Alogoskoufis (1995).
5
A detailed Appendix on the ADF tests is available from the authors on request.
6
The lag length is selected by starting with six lags on every variable, and sequentially testing from
the highest order using an F-test. The same number of lags is obtained by the Akaike Information
Criterion (AIC). All estimations are obtained using PcFiml 9.0 (see Hendry and Doornik, 1997).
7
Table 2, which reports the λi eigenvalues, the λ-max and the trace statistics, and the 5%
critical values. The critical values for these tests are from Osterwald-Lenum (1992).7 The
λ-max statistic rejects no cointegration in favour of r = 1 cointegrating relationship,
whereas the trace statistic supports the existence of r = 2 cointegrating vectors. In what
follows, we assume two cointegrating vectors as suggested by our priors based on the
theoretical model described in the previous section. Table 3 reports the normalised
unrestricted β eigenvectors and the corresponding α coefficients.
Table 1: Diagnostic statistics
Statistic
lT
(pT – p)
(w – pT)
vT
F ar (5,101)
0.504 [0.773]
0.662 [0.653]
0.327 [0.896]
1.585 [0.171]
F arch (4,98)
0.102 [0.982]
0.049 [0.995]
1.327 [0.266]
0.362 [0.835]
χ2 nd (2)
2.553 [0.279]
2.725 [0.256]
2.197 [0.333]
1.567 [0.457]
F het (42,63)
0.873 [0.677]
0.859 [0.697]
0.460 [0.998]
0.760 [0.827]
Notes: F ar is the Lagrange Multiplier F-test for residual serial correlation of up to fifth order. F arch is the
fourth order Autoregressive Conditional Heteroscedasticity F-test. χ2 nd is a Chi-square test for normality. F
het is an F test for heteroscedasticity. Numbers in parentheses indicate the degrees of freedom of the
test statistics. Numbers in square brackets are the probability values of the test statistics.
Table 2: Eigenvalues, test statistics, and critical values
λi
λ-max
H0
H1
Stat.
95%
0.332
r=0
r=1
55.35
0.129
r≤1
r=2
0.053
r≤2
0.025
r≤3
λ-trace
H0
H1
Stat.
95%
27.07
r=0
r≥1
85.28
47.21
18.98
20.97
r≤1
r≥2
29.93
29.68
r=3
7.51
14.07
r≤2
r≥3
10.95
15.41
r=4
3.44
3.76
r≤3
r=4
3.44
3.76
Notes: r denotes the number of cointegration vectors. The critical values of the λ-max and λ-trace
statistics are taken from Osterwald-Lenum (1992).
7
In a cointegrating vector which involves real output, inclusion of a linear deterministic trend may
capture slowly evolving stocks of knowledge as well as human and physical capital (Hendry and
Doornik, 1994). In our case, however, the deterministic trend turned out to be insignificant and was
therefore excluded from the subsequent analysis.
8
Table 3: Estimated cointegrating vectors (β) and weights (α)
β1
β2
α1
α2
lT
1.000
-1.928
-0.001
0.091
(pT – p)
-5.590
1.000
0.026
0.004
(w – pT)
-1.777
-0.239
-0.010
-0.020
vT
-0.613
0.735
0.004
0.004
5.821
0.442
-
-
(pO – pT)*
To uniquely identify the two cointegrating relationships, given their possible
interpretations as the long-run labour demand and price competitiveness equations (11)
and (12) above, we excluded pT – p from the first vector, and lT from the second one.
Then we tested the validity of four over-identifying restrictions. In particular, we
imposed a zero effect from (pO – pT)*, and equal but opposite effects from w – pT and vT
in the first vector. In the second vector, we excluded vT, and restricted the coefficient
on w – pT to be equal to the coefficient on (pO – pT)* plus one.8
It is worth noticing that in the first cointegrating vector, the coefficient on (pO –
pT)* does not have the expected negative sign; however, it is not statistically different
from zero. At the same time, we decided not to impose a unit coefficient on vT as this
restriction was rejected at the 5% significance level. The Likelihood Ratio (LR) test
statistic for testing all four over-identifying restrictions discussed above, is distributed as a
χ2(4) under the null, giving a value of 7.82 which is insignificant at the 5% level (i.e. the
corresponding p-value = 0.098). Imposing the restrictions discussed above, yields the
following cointegrating vectors:
lT = 0.421 vT − 0.421 ( w − pT ) ,
8
In the theoretical equation (12), the coefficient on w – pT is equal to the coefficient on (pO – pT)*
minus one. However, to test this restriction within the cointegrating space, it must be formulated such
that the coefficient on w – pT is equal to the coefficient on (pO – pT)* plus one.
9
and
( pT − p) = −0.371 ( w − pT ) + 0.629 ( pO − pT ) *
The first vector is interpreted as a long-run labour demand equation, with the
coefficient on output estimated at 0.421 and σ also estimated at 0.421. The theoretical
model suggested a proportional relationship between employment and output; however,
our finding of a lower effect from vT on lT implies that labour may be treated as a quasifixed factor of production, due to expectations about the permanent or temporary nature of
changes in production and requirements specific to a firm (Oi, 1962).
The second cointegrating vector can be interpreted as a price competitiveness
equation, with the share of tradeables in total output estimated at τ = 62.9%. Further, τ
measures the effect of oil price increases on the real exchange rate. In particular, since
Greece is an oil importer, an increase in the world price of oil has the effect of depreciating
the real exchange rate in the long-run, and vice versa. It should be noted that there are other
possible determinants of the real exchange rate both in the long and the short run (e.g. the
level and composition of government expenditure, import tariffs, and the extent of capital
controls in the economy, among others; see Edwards, 1989). However, the formulation of
a complete real exchange rate determination model is beyond the scope of this paper.
Table 4 reports the growth equations for ∆lT, ∆(pT – p), ∆(w – pT), and ∆vT using
Full Information Maximum Likelihood (FIML) estimates. Most insignificant variables have
been dropped based on the χ2 version of the LR-test of over-identifying restrictions. Under
the null hypothesis of valid over-identifying restrictions, the test gives χ2(45) = 19.19 which
is insignificant (p-value = 0.99).
10
Table 4: FIML estimates of the error correction model
Variable
∆lT
∆(pT – p)
∆(w – pT)
Coeff.
∆ lT t
∆ lT t-1
∆ lT t-2
∆ lT t-3
∆ lT t-4
∆(pT – p) t-1
∆(pT – p) t-2
∆(pT – p) t-3
∆(pT – p) t-4
∆(w – pT) t
∆(w – pT) t-1
∆(w – pT) t-3
∆(w – pT) t-4
∆ vT t-1
∆ vT t-2
∆ vT t-3
∆ vT t-4
∆(pO – pT)* t
∆(pO – pT)* t-1
∆(pO – pT)* t-2
∆(pO – pT)* t-3
∆(pO – pT)* t-4
CV1 t-1
CV2 t-1
DVAL
σ
F ar (5,104)
F arch (4,101)
χ2 nd (2)
F het (47,61)
S. E. Coeff.
-0.158
-0.465
-0.410
-0.157
0.139
0.091
0.197
0.100
0.195
0.134
0.030
S. E. Coeff.
S. E.
0.182
0.104
0.215
0.108
0.101
0.084
0.092
0.089
0.076
0.130
0.092
0.316
0.136
-0.158
-0.138
0.095 -0.293
0.069 -0.142
0.075
0.091 0.205
0.127
0.086
0.168
0.086
0.305
-0.165
-0.200
0.056 0.228
0.058
0.057
0.054
0.168
0.023 0.074
0.075 -0.260
0.069 0.081
0.071 0.266
0.047 -0.113
-0.075
-0.099
-0.065
0.023 -0.084
0.026 -0.127
0.095
0.070
0.093
0.068 -0.184
0.060 -0.289
0.061 -0.102
0.055 0.510
0.030
0.034
-0.097
0.070
0.076
0.071
0.073
0.050
0.047
0.052
-0.127
0.057
S. E. Coeff.
∆ vT
0.118
0.033
0.022
0.020
0.034
0.022 -0.059 0.020
0.111 0.018
Diagnostic Tests
0.023
0.024
1.971 [0.089] 1.361 [0.245]
0.157 [0.959] 0.037 [0.997]
3.823 [0.148] 5.636 [0.060]
0.783 [0.807] 1.006 [0.486]
-0.038
0.027
0.042
-0.172
0.030
0.023
0.032
0.983 [0.432]
1.274 [0.285]
4.615 [0.100]
0.364 [1.000]
0.033
2.894 [0.017]
0.032 [0.998]
5.990 [0.050]
0.683 [0.913]
Notes:
All regressions also include a constant and centred seasonal dummies. σ is the standard error of the
regression.
The diagnostic tests are defined in the notes of Table 1, and the numbers in square brackets denote
probability values.
CV1 = lT − 0.421vT + 0.421(w – pT), and CV2 = (pT – p) + 0.371(w – pT) – 0.629(pO – pT)*.
11
The equation for ∆lT reveals that employment in the manufacturing sector depends
negatively on the first cointegrating vector (i.e. CV1) with an estimated coefficient equal to
-0.126, which suggests that approximately 50% of the adjustment takes place within a
year.9 Further, the estimated coefficient on the second cointegrating vector (i.e. CV2) is
positive and significant. This suggests that when the real exchange rate is above its long-run
equilibrium level (i.e. the real exchange rate is undervalued) labour in the manufacturing
sector rises; put another way, firms increase their demand for labour following an
improvement in competitiveness.
The equation for ∆(pT – p) provides some insights on the determinants of the real
exchange rate in the short-run, which depreciates following an increase in the world price of
oil. In addition, the coefficient on CV2 has the expected negative sign and suggests that
approximately 20% of the discrepancy between the real exchange rate and its long-run
equilibrium level (real exchange rate misalignment) is corrected within a year.
The equation for ∆(w – pT) depends on its own past values as well as those of the
other variables in the system. Neither of the two error correction terms is found to affect
wage growth in the manufacturing sector. This result can be explained by the institutional
characteristics of the Greek economy, namely the fact that wage decisions take place within
the government sector, which sets the main guidelines for wage increases in the
manufacturing sector (Katseli, 1990).
The equation for ∆vT reveals that output growth in the manufacturing sector
depends positively on employment and the relative price of traded goods (that is, the real
9
Using annual data, Milas (1997) found an adjustment coefficient of -0.093. In that model, however, the
dependent variable was number of employees, whereas here the regressand refers to total hours worked by
persons. This difference is due to the fact that firms find it easier to adjust hours worked rather than
the number of employees, because of strict labour legislation which limits lay-offs without government
authorisation to a maximum of 2% of the firm’s monthly workforce.
12
exchange rate), and negatively on oil prices. Interestingly, we also find a weak positive
effect from the discrepancy between the real exchange rate and its long-run equilibrium
level on ∆vT. This result suggests that if the real exchange rate is undervalued (i.e.
CV2 > 0), the economy becomes more competitive. As a result, employment in the
tradeable sector rises (see the positive effect from CV2 in the ∆lT equation) and at the same
time, production of tradeable goods expands.
The bottom panel of Table 4 reports the diagnostic tests for all four equations. As
can be seen, the equations pass the LM test for residual serial correlation of up to fifth order
(although in the case of ∆vT the test is rejected at the two per cent significance level),
Engle’s LM test for ARCH of up to fourth order, the test for normality, and White’s test
for heteroscedasticity. The stability of the system is analysed using recursive FIML
estimates. The plots of the 1-step residuals ± 2*standard errors (SE) suggest reasonable
parameter constancy for all four equations. Stability is also confirmed by the sequences of
one-step forecast tests (1↑ step Chow test), break-point F-tests (N↓ step Chow test), and
forecast F-tests (N↑ step Chow tests), which are all insignificant at the one per cent level
(see Figure 1).
4. Concluding remarks
In this paper we have introduced a theoretical labour market model for the
tradeable sector of a small open economy. Using quarterly data for Greece, the model
has been estimated following the Johansen (1988, 1995) cointegrating approach, which
provides a test for the existence of, say, r cointegrating vectors that are chosen on
purely statistical terms. In order to translate these vectors to meaningful economic
relationships, we need to take into account the a priori information provided by
13
economic theory. The results of our modelling exercise provide some evidence in
favour of the two long-run relationships derived from the theoretical model. Indeed,
our findings support the existence of two cointegrating vectors. The first one can be
interpreted as a long-run labour demand equation, whereas the second one can be thought
of as a price competitiveness equation.
Turning to the short-run dynamics of the model, labour decisions in the tradeable
sector are found to be affected by past disequilibria in the labour market and, perhaps
more interestingly, the discrepancy of the real exchange rate from its implied long-run
equilibrium relationship. The latter result has an important policy implication as it
shows that temporary misalignments of the real exchange rate will affect employment
and output decisions in the tradeables sector. In a future modelling exercise, it would
be interesting to extend the analysis by incorporating some factors (such us the
decomposition of government expenditure into tradeables and non-tradeables) that may
also affect the behaviour of the real exchange rate.
14
Figure 1. Parameter constancy tests.
Residuals of ∆lT ± 2 S.E.
1↑ Chow tests
1% crit.
1
.05
.75
.5
0
.25
-.05
1980
1
1985
1990
1995
1980
1
.75
.75
.5
.5
1990
1995
1985
1990
1995
.25
.25
1980
1985
1985
N↑ Chows
1990
1995
1980
1% crit.
N↓ Chows
Residuals of ∆(pT – p) ± 2 S.E. 1↑ Chow tests
1% crit.
1% crit.
1
.05
.75
0
.5
.25
-.05
1980
1
1985
1990
1995
1980
1
.75
.75
.5
.5
1990
1995
1985
1990
1995
.25
.25
1980
1985
1985
N↑ Chows
1990
1% crit.
1995
1980
N↓ Chows
1% crit.
15
Figure 1 (Continued) Parameter constancy tests.
Residuals of ∆(w – pT) ± 2 S.E. 1↑ Chow tests
1% crit.
1
.05
.75
0
.5
-.05
.25
1980
1
1985
1990
1995
1980
1
.75
.75
.5
.5
1990
1995
1985
1990
1995
.25
.25
1980
1985
1985
N↑ Chows
1990
1995
1980
1% crit.
N↓ Chows
Residuals of ∆vT ± 2 S.E.
1↑ Chow tests
1% crit.
1% crit.
1
.05
.75
0
.5
-.05
.25
1980
1
1985
1990
1995
1980
1
.75
.75
.5
.5
1990
1995
1985
1990
1995
.25
.25
1980
1985
1985
N↑ Chows
1990
1% crit.
1995
1980
N↓ Chows
1% crit.
16
Data appendix
lT: Total hours worked by persons in manufacturing (1990=100). Source: Main
Economic Indicators, OECD.
vT: Index of manufacturing production (1990=100). Source: Main Economic
Indicators, OECD.
pT: Price of tradeables (in domestic currency) constructed as pT = pT* + e, where pT* is
the dollar unit value index of industrial countries imports (1990=100), and e is
the average drachmae / dollar exchange rate. Source: International Financial
Statistics, IMF.
p: Consumer price index (1990=100), which excludes food and rent. Source: Main
Economic Indicators, OECD.
w: Index of hourly earnings in manufacturing (1990=100). Source: International
Financial Statistics, IMF.
pO: Oil price in domestic currency, constructed as pO = pO*+ e, where pO* is the average
oil price ($ per barrel) and e as above. Source: International Financial Statistics,
IMF.
All variables are in logs.
References
Alogoskoufis, G., 1990. Competitiveness, wage adjustment and macroeconomic policy
in the dependent economy: the case of Greece. Greek Economic Review 12, 1557.
Alogoskoufis, G., 1995. The two faces of Janus: institutions, policy regimes and
macroeconomic performance in Greece. Economic Policy 20, 149-194.
17
Bean, C.R., 1994. European unemployment: a survey. Journal of Economic Literature
32, 573-619.
Bruno, M., Sachs, J.D., 1985. Economics of Worldwide Stagflation, Blackwell,
Oxford.
Edwards, S., 1989. Real Exchange Rates, Devaluation, and Adjustment: Exchange
Rate Policy in Developing Countries, MIT Press, Cambridge, MA.
Ericsson, N.R., Hendry, D.F., Tran, H.A., 1994. Cointegration, seasonality,
encompassing and the demand for money in the United Kingdom. In:
Hargreaves, C., (Ed.) Non Stationary Time Series Analysis and Cointegration,
Oxford University Press, Oxford.
Hendry, D.F., 1995. Dynamic Econometrics, Oxford University Press, Oxford.
Hendry, D.F., Doornik, J.A., 1994. Modelling linear dynamic econometric systems.
Scottish Journal of Political Economy 41, 1-33.
Hendry, D.F., Doornik, J.A., 1997. Modelling dynamic systems using PcFiml 9.0 for
Windows, International Thomson Business Press.
Johansen, S., 1988. Statistical analysis of cointegration vectors. Journal of Economic
Dynamics and Control 12, 231-254.
Johansen,
S.,
1995.
Likelihood-Based
Inference
in
Cointegrated
Vector
Autoregressive Models, Oxford University Press, Oxford.
Katseli, L.T., 1990. Structural Adjustment of the Greek Economy. CEPR Discussion
Paper 374.
Layard, R., Nickell, S., Jackman, R., 1991. Unemployment: Macroeconomic
Performance and the Labour Market, Oxford University Press, Oxford.
18
Milas, C., 1997. Labour market decisions and Greek manufacturing competitiveness.
Journal of Policy Modeling, forthcoming.
Oi, W., 1962, Labour as a quasi-fixed factor of production. Journal of Political
Economy 70, 538-555.
Osterwald-Lenum, M., 1992. A note with quantiles of the asymptotic distribution of
the maximum likelihood cointegration rank test statistics. Oxford Bulletin of
Economics and Statistics 54, 461-472.
Wickens, M.R., 1996. Interpreting cointegrating vectors and common stochastic
trends. Journal of Econometrics 74, 255-271.